Bayes-optimal inverse halftoning and statistical mechanics of the Q-Ising model

TL;DR

This paper develops a Bayesian inference method based on the Q-Ising model and statistical mechanics to improve inverse halftoning, successfully restoring grayscale images from halftoned versions with minimal error.

Contribution

It introduces a novel Bayesian approach using the Q-Ising model for inverse halftoning, identifying conditions for optimal image restoration.

Findings

Bayes-optimal solutions minimize mean-square error in inverse halftoning.

Monte Carlo simulations confirm the theoretical analysis.

The method effectively restores grayscale images like Lenna from halftoned versions.

Abstract

On the basis of statistical mechanics of the Q-Ising model, we formulate the Bayesian inference to the problem of inverse halftoning, which is the inverse process of representing gray-scales in images by means of black and white dots. Using Monte Carlo simulations, we investigate statistical properties of the inverse process, especially, we reveal the condition of the Bayes-optimal solution for which the mean-square error takes its minimum. The numerical result is qualitatively confirmed by analysis of the infinite-range model. As demonstrations of our approach, we apply the method to retrieve a grayscale image, such as standard image `Lenna', from the halftoned version. We find that the Bayes-optimal solution gives a fine restored grayscale image which is very close to the original.